AI BriefingLlamaIndex publishes Document Parsing Protocol 0.1 with 140 rules for standardized output
·2026.10.28 09:00
Key point
The working draft defines 140 rules for handling tables, forms, and diagrams to resolve inconsistencies between different parsers.
Details
- LlamaIndex has published working draft 0.1 of the Document Parsing Protocol (DPP), a proposed definition for document parsing that serves as a shared answer key.
- The protocol includes 140 rules covering text, tables, forms, charts, and diagrams, each accompanied by example pages and expected outputs.
- DPP addresses structural ambiguities, such as multi-row table headers, by specifying that HTML tables be used instead of Markdown to preserve column spans and hierarchy.
- The protocol distinguishes between acceptable variations, such as the presence or absence of and wrappers, and defines handling for hidden text, illegible content, and diagram conversion to Mermaid.
- DPP is available at parseprotocol.com, where users can test the protocol using Claude or ChatGPT by uploading specific skills or reference files.
- The project is open for community input, with issues and pull requests accepted on GitHub to refine the rules based on real-world document failures.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
LlamaIndex launches OpenDocRouter to unify document parsing across frontier and open-source models